Recursive SVD-Based Least Squares Algorithm with Forgetting Factors for Neuro-fuzzy Modeling

Lee and Ouyang [3], [5] proposed a recursive SVD-based least squares estimator (RSVDLSE) for parameter identification of TSK-type neuro-fuzzy system modeling. Besides, Ouyang et al. [4] combined it into a self-constructing rule generation method for structure identification. However, RSVDLSE and its related applications are suitable for solving modeling problems of time-invariant systems, instead of time-varying systems. To extend RSVDLSE for both kinds of systems, we add a term of forgetting factors into the objective function and derive its corresponding estimator, resulting in a recursive SVD-based least squares estimator with forgetting factors (RSVDLSEF). Also, RSVDLSE is a special case of RSVDLSEF. Experimental results have shown that RSVDLSEF tracks the time-varying parameters and system outputs very well and produces smaller estimation errors than RSVDLSE.